Confident Privacy Decision-Making in IoT Environments

Abstract

Researchers are building Internet of Things (IoT) systems that aim to raise users’ privacy awareness, so that these users can make informed privacy decisions. However, there is a lack of empirical research on the practical implications of informed privacy decision-making in IoT. To gain deeper insights into this question, we conducted an online study (N = 488) of people’s privacy decision-making as well as their levels of privacy awareness toward diverse IoT service scenarios. Statistical analysis on the collected data confirmed that people who are well aware of potential privacy risks in a scenario tend to make more conservative and confident privacy decisions. Machine learning (ML) experiments also revealed that individuals overall privacy awareness is the most important feature when predicting their privacy decisions. We verified that ML models trained on privacy decisions made with confidence can produce highly accurate privacy recommendations for users (area under the ROC curve (AUC) of 87%). Based on these findings, we propose functional requirements for privacy-aware systems to facilitate well-informed privacy decision-making in IoT, which results in conservative and confident decisions that enjoy high consistency.